How to Add an AI Chatbot to WhatsApp Step by Step (2026)

Updated: June 4, 2026

A customer writes at 10:30 pm asking about availability, sends a voice note describing their problem or pastes a URL to compare options. If nobody replies until the next day, that lead is already talking to your competitor. Putting conversational AI on WhatsApp solves that moment: it answers, qualifies and proposes the next step without your team being glued to the phone. This guide covers the four steps to get it right from day one.

What you need before you start

A conversational WhatsApp AI is an intelligent layer that runs on the WhatsApp Business API, Meta's official channel for automating messages at scale. For an SMB, it makes sense when the volume of enquiries exceeds what the team can handle during business hours, and it cuts the risk of losing leads after hours.

The channel justifies it. According to the ONTSI report La Sociedad Digital en España 2024, over 90% of Internet users in Spain use WhatsApp daily, and the CNMC put daily instant-messaging use at close to 70% of the population. Before you start, make sure you have:

• A phone number that is not registered on WhatsApp Personal or the WhatsApp Business app.
• A verified business account in Meta Business Suite.
• Clarity on what you want the AI to handle: answering questions, qualifying leads, booking, recovering conversations, or a combination.
• The person who will receive escalations when the AI should not make the call.

Step 1: Register for the WhatsApp Business API

Registering a number on the WhatsApp Business API is the first technical step to automating replies. Without that official connection, any AI solution is limited to the WhatsApp Business app, which does not allow advanced flows or comply with the 24-hour window using approved templates, per Meta's documentation.

1. Request access to the WhatsApp Business API through Meta Business Suite. You need a verified business ID and a clean phone number with no WhatsApp usage history.
2. Verify the number with Meta. Meta sends a verification code to the selected number. Once verified, the number is linked to the API and can no longer be used from the regular WhatsApp app.
3. Create and get message templates approved for proactive communications. Templates are predefined messages that Meta reviews and approves. They are used to reopen conversations outside the 24-hour window. Inside the window, the AI can reply freely.

Meta's official WhatsApp Business Platform documentation describes these requirements in detail. Working with a provider that manages the registration shortens the wait and avoids verification errors.

Step 2: Choose and configure the AI

Choosing the AI that handles the conversations defines the real scope of the system. For your business, what matters is that it understands voice, images and URLs, keeps context across several turns, and can escalate the case to a person when the decision requires human judgement.

Tubot runs on the WhatsApp Business API and adds four capabilities that expand what can be resolved in the same thread:

• VOICE: the lead sends an audio, Tubot transcribes it, detects the language and can reply with a voice note when the answer does not include a URL.
• VISION: the lead sends a photo and Tubot analyzes it within the conversation.
• WEB: the lead pastes a URL and Tubot reads the content to answer about that specific case.
• FOLLOW UP: after 4 hours of silence, Tubot sends a voice note with a specific observation about the case to recover the conversation.

During initial setup, load your business information: hours, services, published prices, escalation rules and answers to the most frequent questions. You can see how different industries have configured their flows in the articles on WhatsApp AI for restaurants and WhatsApp AI for ecommerce.

Step 3: Define flows and escalation rules

Defining flows means deciding which questions the AI closes and which cases escalate to the team. Without those rules the system replies but does not move forward: it half-qualifies, does not book, does not recover leads. With them, every conversation has a clear next step for the lead and for the team.

A minimum viable flow covers three moments:

1. First message: the AI greets, understands the intent and answers the question or asks for the minimum data needed to do so.
2. Qualification: it collects area, budget, urgency or availability depending on the industry, and proposes a concrete action such as booking, quoting or confirming.
3. Escalation or close: if the decision requires professional judgement, the AI hands over with the context prepared; if not, it closes the conversation with a clear next step for the lead.

Also set the limits: which types of enquiry the AI should not answer and which signals trigger an immediate handover to a person. This pattern matches the one we apply in the guide on automating orders via WhatsApp. To review your case, Tubot's contact page is the starting point.

How long until the WhatsApp AI is up and running?

Time to launch depends on the complexity of the flows, the integrations with calendars or a CRM, and how fast Meta verifies the number. A first, well-scoped flow can be live in a few days if the business information is well organized.

The most time-consuming steps are number verification with Meta and approval of the first templates. Configuring the AI and the initial flows goes faster if the escalation criteria are defined from the start. Tubot runs in production at tubot.es and the free audit helps estimate the real timeline for each specific business.

What it looks like in practice

A well-configured conversation asks for little data and returns a concrete action. This example shows the four steps applied to a services business: the lead arrives via WhatsApp, the AI qualifies and books the visit without the team stepping in.

Metrics to validate that it works

A well-configured WhatsApp AI should move business numbers, not just message counts. The useful metrics in the first 30 days show whether you reply sooner, qualify better and turn the conversation into an action the team can measure.

• First response time: from the lead's first message to the AI's first useful reply.
• Percentage of conversations with a concrete action: how many end with an appointment booked, a quote sent or a lead qualified without human intervention.
• Leads recovered by follow-up: conversations that come back after the automatic voice note at 4 hours.
• Correct escalations: cases that reach the team with enough context to act without asking the lead to repeat information.

Is your business still losing leads every night because nobody replies when the customer finally has time to write?

Free audit. Ask for it on WhatsApp now: the team reviews your case and replies in under 30 seconds.